Transcript Click to open

The USC Pmetrics and RightDose software
The ONLY software with integrated pop
modeling, simulation, and maximally precise
dosage
c
c
R Jelliffe, A Schumitzky , D Bayard, R Leary, M Van Guilder, S Goutelle, A Bustad,
A Botnen,
A Zuluaga, J Bartroff, W Yamada, and M Neely
ABSTRACT
,
The Pmetrics
population
modeling
software
is
embedded
in
R,
called
by R, and output into R.
., Laboratory of Applied Pharmacokinetics,
USC Keck School of Medicine, Los
It runs on PC’s and Macs. Minimal experience with R is required, but the user has all the
power of R for further analyses and displays, for example. Libraries of many models areAngeles CA
available. In addition, differential equations may also be used to describe large models of
multiple drugs, with interactions, and with multiple outputs and effects. Analytic solutions may
also be used if applicable. The model is compiled with GFortran. Runs are made with simple
R commands. Routines for checking data and displaying results are provided. Likelihoods are
exact. Behavior is statistically consistent – studying more subjects yields parameter estimates
closer to the true ones. Stochastic convergence is as good as theory predicts. Parameter
estimates are precise [1].The software is available freely for research uses. In addition,
MM maximally precise stepwise lido infusion regimen: Predicted
prototype new nonparametric Bayesian (NPB) software has been developed. Standard errors
Lidocaine stepwise infusion regimen based on Parameter MEANS:
response of full 81 point lidocaine population model. Most precise
of parameter estimates and rigorous Bayesian credibility intervals are now available. This
Predicted response of full 81 point lidocaine population model.
regimen. Target = 3ug/ml
Target = 3ug/ml
work, presented at this meeting, is progressing.
The RightDose clinical software [2] uses Pmetrics population models, currently for a 3
compartment linear system, and develops multiple model (MM) dosage regimens to hit
desired targets with minimum expected weighted squared error, thus providing maximally
precise dosage regimens for patient care. If needed, hybrid MAP and NP Bayesian posteriors Support point values don’t change. Compute Bayesian posterior
provide maximum safety with more support points and more precise dosage regimens. In
probability of each support point, given the patient’s data. Problem:
addition, the interacting multiple model (IMM) sequential Bayesian analysis when model
will not reach out beyond pop parameter ranges. May miss unusual
parameter distributions are changing during the period of data analysis [[3] has been
upgraded by using the hybrid analysis in advance to provide more support points than were patient.
present in the original population model, again for more capable Bayesian parameter
distributions and more informed dosage regimens than were available before. This work was
also presented at this meeting. IMM has tracked drug behavior better than other methods in
tart with MAP Bayesian estimate. Reaches out pop ranges to unusual
unstable post surgical cardiac patients [4]. In all the software, creatinine clearance is
estimated in either stable or changing clinical situations, based on analyzing pairwise serum patients. Add even more support points nearby, to augment original
creatinine values, age, gender, height, weight, muscle mass, and dialysis status [5].
pop model. Then MM Bayesian analysis. More flexible, informed, safer.
References:
Bustad A, Terziivanov D, Leary R, Port R, Schumitzky A, and Jelliffe R: Parametric and
Nonparametric Population Methods: Their Comparative Performance in Analysing a Clinical
Data set and Two Monte Carlo Simulation Studies. Clin. Pharmacokinet. 45: 365-383, 2006.
Jelliffe R, Schumitzky A, Bayard D, Milman M, Van Guilder M, Wang X, Jiang F, Barbaut X,
Limitation of all other current Bayesian methods - find only 1 set of
and Maire P: Model-Based, Goal-Oriented, Individualized Drug Therapy: Linkage of
population Modeling, New “Multiple Model” Dosage Design, Bayesian Feedback, and
fixed parameter values which fit the data.IMM - Relax this assumption.
Individualized Target Goals. Clin. Pharmacokinet. 34: 57-77, 1998.
Let “true patient” change during data analysis if more likely to do so.
Bayard D and Jelliffe R: A Bayesian Approach to Tracking Patients having Changing
More available support points in new version. More flexible now.
Pharmacokinetic Parameters. J. Pharmacokin. Pharmacodyn. 31: 75-107, 2004.
W g tA vg
MacDonald I, Staatz C, Jelliffe R, and Thomson A: Evaluation and Comparison of Simple
Multiple Model, Richer Data Multiple Model, and Sequential Interacting Multiple Model (IMM)
1 2 .5
Bayesian Analyses of Gentamicin and Vancomycin data collected from Patients undergoing
1 0 .0
Cardiothoracic Surgery. Ther. Drug Monit. 30: 67-74, 2008.
Jelliffe R: Estimation of Creatinine Clearance in patients with Unstable Renal Function, without
7 .5
4
MM BAYESIAN ANALYSIS.
NEW! IMPROVED HYBRID BAYESIAN ANALYSIS
S
NEW! IMPROVED INTERACTING MULTIPLE MODEL (IMM)
C o n c e n tr a tio n in c e n tr a l c o m p a r tm e n t [u g /m L ] C o n c e n tr a tio n in c e n tr a l c o m p a r tm e n t [u g /m L ]
BAYESIAN ANALYSIS FOR UNSTABLE PATIENTS
a Urine Specimen. Am, J, Nephrology, 22: 320-324, 2002.
: 3200-324, 2002.
NONPARAMETRIC POPULATION MODELS
•Get the entire ML distribution, a Discrete Joint Density: one
parameter set per subject, + its probability.
•Shape of distribution determined only by the data itself.
•Multiple individual models, up to one model set per subject.
•Can discover, locate, unsuspected subpopulations.
•Behavior statistically consistent. Study more subjects, better
results.
•The multiple models permit multiple predictions.
•Can optimize precision of goal achievement by a MM dosage
regimen.
•Use IIV +/or assay SD, stated ranges.
•Computes environmental noise.
•Bootstrap, for confidence limits, significance tests.
5
5 .0
2
2 .5
6
3
1
7
8
9
1 0
0 .0
0
1 0 0
2 0 0
3 0 0
4 0 0
T i m e [ h o u rs ]
5 0 0
6 0 0
7 0 0
MM Bayesian updating – only fair tracking in unstable patient
W g tA vg
2 0 .0
1 7 .5
1 5 .0
1 2 .5
1 0 .0
4
7 .5
5
5 .0
2
6
2 .5
1
3
7
8
9
1 0
0 .0
0
1 0 0
2 0 0
3 0 0
4 0 0
T i m e [ h o u rs ]
5 0 0
6 0 0
7 0 0
IMM: interacting sequential MM Bayesian analysis – best tracking
NEW! Nonparametric Bayesian Population Modeling
We developed a prototype nonparametric Bayesian pop
MULTIPLE MODEL (MM) DOSAGE DESIGN modeling method based on stick breaking to make the prior. It
Use a prior with discrete multiple models - an NPEM or NPAG
has close agreement with Pmetrics (NPAG) but also obtains
model.
Give a candidate regimen to each model. Predict results with SE’s of parameters and rigorous Bayesian credibility intervals.
each model. Compute weighted squared error of failure to hit
target goal
Find the regimen hitting target with minimal weighted squared
error.
This is multiple model (MM) dosage design – the IMPORTANT
clinical reason for using nonparametric population PK models.
NEW! Runs on IPads, IPhones!
Uses these devices as virtual machines to run
RightDose on PC’s.